HAL Id: tel-01784063
https://tel.archives-ouvertes.fr/tel-01784063
Submitted on 3 May 2018HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Ta�N� thin films for the photoelectrolysis of water
Martin Rudolph
To cite this version:
Martin Rudolph. Synthesis and characterization of magnetron-sputtered Ta�N� thin films for the photoelectrolysis of water. Plasma Physics [physics.plasm-ph]. Université Paris Saclay (COmUE), 2017. English. �NNT : 2017SACLS094�. �tel-01784063�
NNT : 2017SACLS094
T
HESE DE DOCTORAT
DE
L’UNIVERSITÉ PARIS-SACLAY
PRÉPARÉE À
L’UNIVERSITÉ PARIS-SUD
E
COLED
OCTORALE N°572
Ondes et Matière
Spécialité de doctorat : Physique des plasmas
Par
M. Martin Rudolph
Synthesis and characterization of magnetron-sputtered Ta
3N
5thin films for the photoelectrolysis of water
Thèse présentée et soutenue à Orsay, le 2 Mai 2017 :
Composition du Jury :
M. Ian Vickridge Directeur de la recherche,
Université Pierre et Marie Curie, Paris
Président
M. Jón Tómas Guðmundsson Professeur
Háskóli Íslands, Reykjavik et Kungliga Tekniska Högskolan, Stockholm
Rapporteur
Mme. Angélique Bousquet Maitre de conférences,
Université Blaise Pascal, Clermont-Ferrand
Rapporteur
M. Philippe Lecoeur Professeur,
Université Paris-Sud, Orsay
Examinateur
Mme. Marie-Christine Hugon Maitre de conférences,
Université Paris-Sud
Directrice de thèse
Mme. Brigitte Bouchet-Fabre Chargée de la recherche,
CEA Saclay
Co-directrice de thèse
Yes, my friends, I believe that water will one day be employed as fuel, that hydrogen and oxygen which constitute it, used singly or together, will furnish an inexhaustible source of heat and light, of an intensity of which coal is not capable.
Jules Verne, The Mysterious Island
Acknowledgments
First of all, I would to like to express my sincerest gratitude to my advisor Marie-Christine Hugon who has shared with me her immense knowledge on various measurement techniques in materials science. I have appreciated all the scientific and technical discussions we had in the last three years out of which new ideas continuously emerged. And I am particularly grateful for having put me in contact with so many researchers from diverse fields to discuss my questions. I strongly believe that this has been a key to the success of such an interdisciplinary study.
Furthermore, I am very grateful to Tiberiu Minea for having accommodated me in the group “Théorie et Modélisation des Plasmas - Décharges et Surface” at the Laboratoire de Physique des gaz et des plasmas (LPGP) and for the many discussions we had around the physics of discharges. Despite his responsibilities as the head of the laboratory later on, it has always been possible to discuss recent results and his advice for simulations and plasma measurements has been very supportive.
Thanks as well to my co-advisor Brigitte Bouchet-Fabre. Her enthusiasm for materials and renewable energies has been a steady source of motivation for me.
I would also like to thank the jury members Ian Vickridge, Jón Tómas Guðmundsson, Angélique Bousquet and Philippe Lecoeur, for having read carefully through my manuscript and for their advice in polishing up the final version. I have very much appreciated their questions and comments during my defense.
I would like to thank the French team of the SNON project: José Alvarez, Sofia Gaiaschi and Jean-Paul Kleider from the GeePS laboratory of Central Supélec in particular for the electronic and AFM measurements; and to Dana Stanescu and Hélène Magnan of the CEA Saclay in particular for the photoelectrochemical measurements. I thank all of them, including Nathalie Herlin-Boime from the CEA Saclay, for the discussions on so many different facets of the study.
I am also very grateful to the Romanian team of the project in particular to Lucel Sirghi for inviting me for a stay at his laboratory at the Alexandru Ioan Cuza University. I enjoyed working with his colleagues, Vasile Tiron, Alexandra Demeter, Ilarion Mihaila and Claudiu Costin, particularly on the mass spectrometry, but I have similarly appreciated the discussions on the physics and chemistry of surfaces.
The measurements at the accelerators at the University of Namur, at the CEA Saclay and the Institut des NanoSciences de Paris at the Université Pierre et Marie Curie have certainly been the most captivating experiments. I am grateful to Jean-Jacques Ganem, Emrick Briand, Hicham Kodja and Tijani Tabarrant for their support with the measurements and the analysis, for sharing their knowledge on nuclear reactions and for the continuous effort to get the most out of the accelerators. With a special mention to Eddy Foy from the CEA Saclay, for having introduced me to the fascinating field of X-ray diffraction and in particular for the countless diffractograms we recorded.
I have not forgotten the day we saw the Ta3N5 phase for the first time.
I am very grateful to Martine Wery for the effort she made to accommodate meetings into her tight schedule to share her vast knowledge on electrochemistry with me.
I am also grateful to Séverine Le Moal from the Institut des Sciences Moléculaires d’Orsay, Diana Dragoe from the Institut de Chimie Moléculare et des Matériaux d’Orsay and Jocelyne Leroy from the CEA Saclay for the XPS measurements and the assistance with fitting and interpreting the
spectra. I have learnt a lot from them. Thanks as well to Patrick Bonnaillie from the CEA Saclay for his incredible skills in taking SEM images.
During my thesis, I have asked many little favors to many people at the LPGP. I am grateful to Joel Baudet for having put my unskilled technical drawings into reality. Thanks to Gérard Bauville for having pushed the limits of his laser cutter further and further. Thanks to João Santos Sousa for his assistance with the pH measurements and for having let me work in his chemistry lab. Thanks to Stéphane Maingot for always having had the right tools at hand. Furthermore, I would like to thank Michel Fleury, Marc Bisson and Michel Pascanu for their their help with electronics and software. Thanks as well to Olivier Antonin, for having shared with me all the tricks to get the sputter system running reliably and finally, thanks to Charles Ballage for helping me find and repair the leak in the deposition chamber.
The administration of the LPGP has always been very supportive. A big thanks to Nicolas Dessaint and the current and former members of his team, Thierry Devillers, Sophie Granon, Ophélie Molle, Bérénice Saidi and Nathalie Rico-Perez.
I have in particular appreciated the working atmosphere at the LPGP. Thanks to all the PhD students, postdocs and researchers that have contributed to this, Jean-Sébastien, Frédéric, Adrien, Lise, Alexis, Patrick, Thomas, Steven, Benjamin, TJ, Felipe, Daniel, Alexandra, Sébastien and Mathieu.
Finally, I would like to thank my parents and grandparents for their support throughout my studies, my siblings for their encouraging words, and all my friends who have supported me in one way or another in the last years.
1
Content
Content ... 1
1 Context, objectives and outline of this thesis ... 3
2 Introduction and literature review ... 7
The photoelectrolysis of water ... 8
The Ta3N5 crystalline phase ... 19
Magnetron sputtering ... 30
Conclusion ... 51
3 Experimental techniques ... 53
Magnetron sputtering system ... 54
X-ray diffraction ... 57
Ion beam analysis ... 60
Total transmission/reflection spectroscopy ... 66
X-ray photoelectron spectroscopy (XPS)... 73
Optical emission spectroscopy (OES) ... 76
Mass spectrometry (MS) ... 77
Photoelectrochemical measurements ... 79
4 Properties of magnetron-sputtered TaN thin films ... 83
Synthesis and characterization of films by DC and HiPIMS ... 84
Mass spectrometry ... 97
Evaluation of backscattered neutrals ... 106
Conclusion ... 114
5 The role of oxygen in Ta3N5 thin films ... 115
Preamble: The necessity for oxygen and its effects ... 116
Sample cleaning ... 117
Ta3N5 thin film deposition ... 121
Sample characterization ... 122
Discussion ... 136
Conclusion ... 137
6 Improving the crystallinity of Ta3N5 thin films ... 139
Guided-ion DC magnetron sputtering ... 140
Proof of concept ... 146
2
Voltammetry measurements ... 157
Conclusions ... 159
7 Conclusions and future work ... 161
Conclusions ... 161
Future work ... 163
8 Appendix: Abbreviations ... 165
9 Appendix: Résumé ... 167
3
1 Context, objectives and outline of this thesis
The vision of a clean energy supply has challenged scientists and engineers for the last two decades. Its attraction springs from the image of tapping virtually infinite resources of energy with little side effects. But it is also based on more rational grounds. Global warming urges us to substitute our fossil energy sources to limit the greenhouse effect. Moreover, emission of particulate matter from coal-fired power plants, risks of nuclear technologies and the reliance on oil-producing, possibly politically unstable countries, to mention only some points, render renewable energies very attractive.
Possible sources of renewable energies available on our planet are manifold and so are the technologies to convert their energy into a useable form: reactors ferment biomass to methane, the tidal range drives electricity generators and solar collectors heat up water for domestic use. The exploitation of various sources is wanted and contributes to a stable and reliable supply. However, in view of the world demand, only a few sources have the potential to contribute a significant share to the world energy supply: biomass, the wind and the sun. Each of the latter two even supplies sufficient power to cover the world demand (Fig. 1). An energy transition is thus to rely on these sources.
Fig. 1: World power demand and total power available by renewable sources [1].
To approach the objective of having an energy supply from entirely clean sources, the European Union set a target. 20% of the total gross energy consumption should come from renewable energy sources by 2020 [2]. And indeed, renewable energies have gained a significant percentage, which rose from 8.7% in 2005 to 15.2% in 2014 [2]. This growth was mainly carried by the installation of wind turbines and photovoltaic panels (Fig. 2) as these technologies reached maturity and became cheaper over the years [3]. There is no doubt that they will continue to play a key role in pushing the share of renewable energies towards the 20% target in the near future.
In the long term, solutions for two major challenges need to be found, in order to reach even more ambitous goals. Firstly, the production of renewable electricity from the sun and the wind is highly fluctuating on various time-scales due to weather and seasonal changes. This, in combination with the unavailability of storage capacity for large amounts of electricity, has already today let to the absurd phenomena of negative wholesale electricity prices [4], which could impair future
4
investments in wind turbines and photovoltaics. The development of storage capacity is thus a prerequisite for further exploiting the wind and the sun for the production of electricity [5]. Secondly, the high contribution of renewable energies to the total energy consumption is shadowing the fact that the three energy sectors, electricity, heating & cooling and transport differ greatly in their contribution to this single figure. While the share of renewables in the electricity sector was at 25.2% in 2014, the transport sector lagged largely behind with only 5.2% of all consumed fuels coming from renewable sources [2]. Cars, vessels and planes still today rely heavily on fossil fuels and this is likely not to change any time soon, as clean transport technologies are still under development [6]. Investing in today’s mature technologies like wind turbines and photovoltaic panels alone, is thus insufficient for reaching the long-term goal of having all energy come from renewable sources.
Fig. 2: Million ton of oil equivalent (Mtoe) of renewable energies decomposed by their source from 2005 to 2013. From [2].
Therefore, development efforts are focussed on the renewable production of chemical fuels and the storage of energy [7]. One possibility to approach both targets at the same time is to convert renewable electricity into chemical fuels by electrolysis. The advantage is that chemical fuels are more practical when it comes to storage. Hydrogen, for example, has still an unprecedented specific energy density of 141MJ/kg compared to commercially available batteries ranging between 0.09MJ/kg and 2.3MJ/kg [8]. Furthermore, its transport by pipelines or in liquid form on trucks or ships is principally possible [9] [10]. But not only could hydrogen serve as an energy storage alone. It can partially replace natural gas used for heating [11]. It is also considered suitable as a fuel in transport. Already today, hydrogen-fueled vehicles are just entering the market [12] [13].
Today, the clean production of hydrogen is possible by using renewable electricity to electrolyze water into oxygen and hydrogen. This is an energy-consuming process as efficiencies of 50% are reached at tops for electrolyzers [10]. In conjunction with a commercially available photovoltaic generator efficiency of 20%, a system like this results in a solar-to-hydrogen efficiency of 10%.
5 This rather low efficiency presents opportunities to alternative technologies that are still at the beginning of their development and thus lag behind in technology readiness level.
One alternative is the photoelectrolysis of water which is the direct use of photon energy for the splitting of water molecules to produce hydrogen and oxygen. It was discovered by Akira Fujishima
and Kinichi Honda in 1971 [14]. Their setup consisted of a semiconducting TiO2 electrode
connected to a platinum cathode both immersed into a water-based electrolyte. When illuminating the setup with ultraviolet light, they observed the evolution of oxygen and hydrogen on the anode and cathode respectively. This groundbreaking experiment triggered a number of developments in the years to follow but still today, the efficiencies remain low.
The principal challenge is to find and synthesize suitable materials. In particular, the requirements on the position of the valence and conduction band edges, combined with the need for a small band gap, are rarely found. In order to overcome this issue, tandem cells have been developed with multiple layers, that have reached quite high efficiencies of up to 14% [15]. From an application point of view, simple systems would be preferred, though, mainly for cost reasons. Those systems, consisting of a photoactive material and possibly a co-catalyst, however, have only reached efficiencies of around 1% in the past [16] [17].
This thesis focusses on the synthesis of Ta3N5, one of the most promising semiconductors for the
photoelectrolysis of water. It has a band gap of 2.1eV and suitable band positions [18] [19]. For its synthesis, magnetron sputtering is chosen. It is technique widely used in industry that offers a large parameter space for tuning film properties. However, until today, the synthesis of this material by magnetron sputtering is little studied. Only three groups of authors have published results on the sputter-deposition of this material [20] [21] [22]. The possibilities offered by this technique thus still lack a systematic investigation.
This thesis identifies the relevant physical mechanism for the successful deposition of Ta3N5.
Moreover, using that knowledge, the deposition process is optimized to obtain high-quality thin
films of Ta3N5. The focus here lies on the bulk properties, i.e. the deposition process is optimized
to obtain highly crystalline films with large grains. The thesis is structured in the following way:
After the general overview of the subject in this chapter, the second chapter presents a more detailed compendium of the three fields relevant for this thesis. It starts with an introduction to the
photoelectrolysis of water. This is followed by a summary of Ta3N5 properties and a presentation
of its synthesis by conventional ammonolysis. The chapter closes with an introduction to magnetron sputtering covering the effects that are important to understand the results of this thesis.
The third chapter is dedicated to the experimental and computational techniques employed throughout this thesis. It covers a description of the sputter deposition system used, introduces X-ray diffraction, ion beam analysis including Rutherford backscattering and nuclear reaction analysis. This is followed by a subsection on transmission and reflection spectroscopy including the presentation of a code, developed during this thesis to fit the experimental spectra to optical models. Finally, the four remaining subsections are dedicated to techniques that played a minor importance during the course of this thesis. They are X-ray photoelectron spectroscopy, optical emission spectroscopy, mass spectrometry and photoelectrochemical measurements.
The fourth chapter is dedicated to the first results obtained from the deposition of δ-TaN films. This phase serves as a model system to identify relevant mechanisms in sputtering from a Ta cathode. Particular focus is placed here on the size of the grains in these films, as this is the parameter to be
6
modes are investigated. The observations are correlated with measurements of the ion flux onto the substrate and an analysis of the backscattered ion energy distribution. This leads to the
recommendation to use the direct current sputtering mode for the deposition of Ta3N5-crystalline
films.
The fifth chapter analyses the influence of oxygen incorporated into magnetron-sputtered Ta3N5
thin films. After giving a recipe for the deposition of Ta3N5-crystalline films, the material properties
as a function of oxygen incorporation are measured. These measurements comprise X-ray diffraction, the determination of the complex dielectric function from which the band gap is determined and electrical measurements, notably the conductivity and the efficiency-mobility-lifetime product. X-ray photoelectron spectroscopy is done before and after the photoelectrochemical measurements, to observe chemical changes to the surface induced by the photoelectrolysis. The films deposited in this chapter are shown to work as photoanodes, but their crystallinity remains small.
The sixth chapter presents the implementation of a technique to increase the ion flux onto the
growing film. This is shown to increase in particular the flux of N2+ onto the growing film, which
effectively increases the degree of crystallinity as observed on the surface of these films by atomic force microscopy. This allows to determine the elemental composition of Ta-O-N
magnetron-sputtered films. The correlate of this with the appearance of the Ta3N5 phase then yields a stability
range with respect to the incorporation of oxygen into the structure. Photoelectrochemical measurements reveal the advantages of having a high incorporation of oxygen into the lattice.
The last chapter concludes the important findings for the sputter-deposition of Ta3N5 and its
7
2 Introduction and literature review
This chapter gives a general introduction to this study. To understand the motivation behind the thesis, this chapter starts with a basic introduction to the photoelectrolysis of water (Section 2.1).
This is followed by a subsection on the Ta3N5 phase comprising a description of its place within the
Ta-O-N ternary system as well as experimentally and theoretically derived physical and electrochemical properties. Its conventional synthesis is shortly discussed at the end to compare it with the synthesis by magnetron sputtering (Section 2.2). The last subsection is dedicated to a general introduction into the magnetron sputtering technique (Section 2.3). A conclusion terminates the chapter (Section 2.4).
2.1 The photoelectrolysis of water 8
2.1.1 The solid state: metals and semiconductors 8
2.1.2 Electrolyte 11
2.1.3 Interfaces 12
2.1.4 Prerequisites on semiconductors 15
2.1.5 Semiconductors for photoelectrolysis 16
2.1.6 Efficiencies and benchmarks 17
2.2 The Ta3N5 crystalline phase 19
2.2.1 Crystalline phases of the Ta-O-N ternary system 19
2.2.1.1 The Ta-N binary system 19
2.2.1.2 Important phases of the ternary Ta-O-N system 21 2.2.1.3 Formation enthalpies of some selected phases 22
2.2.2 The crystalline Ta3N5 phase 23
2.2.2.1 Ta3N5 structure 23
2.2.2.2 Ta3N5 band gap and dielectric function 24
2.2.2.3 Ta3N5 charge carrier mobility 25
2.2.2.4 Ta3N5 band positions 26
2.2.2.5 Ta3N5 stability in photoelectrolytic water splitting experiments 27
2.2.2.6 Surface morphology 28
2.2.3 Conventional synthesis of Ta3N5 28
2.3 Magnetron sputtering 30
2.3.1 Electrical gas discharges 30
2.3.1.1 Discharge regimes 30
2.3.1.2 The glow discharge tube 31
2.3.2 Magnetron discharge 33
2.3.2.1 Magnetron cathode and substrate 33
2.3.2.2 Magnetron deposition 34
2.3.3 Modes of magnetron operation 44
2.3.4 Magnetron sputtering of the Ta-O-N system 45
2.3.5 Magnetron sputtering of Ta3N5 48
2.3.6 Summary of magnetron-sputtered films from the ternary Ta-O-N system 49
8
The photoelectrolysis of water
The photoelectrolysis of water is the splitting of water molecules into hydrogen and oxygen by using directly the energy of incident photons.
2𝐻𝐻20 + ℎ𝜈𝜈 → 2H2+ 𝑂𝑂2 Eq. 1
The photon in this process provides the energy for the uphill reaction. Its Gibbs free energy change
is ∆G = 238 kJ/mol which, broken down to the production of a single Hatom equals 1.23eV.
As water is transparent to visible light, it cannot directly absorb the incident energy. In the photoelectrolysis of water, a semiconducting electrode is therefore utilized which is in direct contact with a water-based electrolyte. It absorbs incident photons which create electron-hole pairs that are separated in an internal electric field. For n-type semiconductors, holes migrate towards the
electrolyte where they recombine with electrons from HO- in the electrolyte under formation of
oxygen. Electrons migrate in the opposite direction towards a metallic counter-electrode from where
they can transfer onto H2O molecules under the formation of hydrogen molecules. The electric field
and the associated band bending that ensure the charge separation is provided by the space charges at the semiconductor/electrolyte interface.
Fig. 3: Schematics of the photoelectrolysis of water using a Ta3N5 semiconductor (red) and a
counter electrode (grey).
Before entering a more detailed discussion on the physics and chemistry of the photoelectrolysis, the basic concepts used to describe the behavior of electrons in solids, i.e. metals and semiconductors, and aqueous electrolytes are introduced. This is followed by a more detailed discussion on the individual processes at the semiconductor-solution and the metal-solution interface.
2.1.1 The solid state: metals and semiconductors
Contrary to free electrons, electrons in a crystalline solid material are bound in an electrostatic well with a periodic potential formed by the atomic nuclei and their core electrons. In this potential,
9 electrons occupy energy levels that can be grouped into bands, whereas each band has a dispersion relation linking an energy state to a wave vector of the electron wave function. Each of these states, by reason of the symmetry of the crystal, can be occupied by two electrons with opposite wave vectors. A fully occupied band thus cannot contribute to electrical conduction, as an electron propagating into one direction has a counterpart propagating into the opposite direction. Partially filled bands, on the other hand, render a material conductive, and its electrical properties depend exclusively on these partially filled bands [23].
The distribution of electrons among possible energy states follows the Fermi-Dirac statistics which
is a function of the temperature. In the ground state, i.e. at a temperature of 0K, only the lowest
lying levels are occupied up to the Fermi energy. At higher temperatures, the distribution extends beyond the Fermi energy as some electrons have sufficient energy to occupy higher-lying energy states. This occupancy of states is described by the Fermi-Dirac distribution [24]:
𝑓𝑓(𝐸𝐸) = 1
1 + exp (𝐸𝐸 − 𝐸𝐸𝐹𝐹 𝑘𝑘𝐵𝐵𝑇𝑇 )
, Eq. 2
where EF is the Fermi level, that according to above equation is the level at which an energy state
has the probability of 1/2 to be occupied. It is typically measured against the energy of a free electron in vacuum and takes values of a few eV.
Metals
In a metal, the Fermi level lies within a band, which is naturally only partially occupied. Therefore metals are conductive and the electrons can be approximately described as a free electron gas [25]. The depth of the potential well, in which electrons are located is termed the work function, which is the energy required to remove an electron from the Fermi level to the field-free vacuum. The
energy required for that is the sum of |EF| and a contribution from a dipole barrier at the surface
from to lattice relaxation [26] and contaminants [27]. In addition, surface treatments can alter the work function of a material drastically. While polycrystalline gold has a work function of 4.4eV to 4.7eV under ambient atmosphere, sputter cleaning the same material gives it a work function of
between 5.3eV and 5.4eV [28]. The work function is thus only a rough approximation for |EF|.
Semiconductors
In intrinsic semiconductors, the Fermi level lies within a band gap, a gap in energy space without
any electronic states. At the absolute zero, all bands below EF (valence bands) are fully occupied,
while those above EF (conduction bands) are empty. As a consequence, electronic conduction is not
possible and the material remains an insulator. As an example, Fig. 4 shows the band structure of
the metallic δ-TaN phase and the semiconducting Ta3N5 phase. The Fermi level EF in these cases is
10
Fig. 4 : Band structure as calculated by density functional theory (DFT) for the metallic δ-TaN (left) and the semiconducting Ta3N5 phase (right). The Fermi level EF is indicated by a horizontal
line at 0.0eV. M, Γ, X, R indicate crystallographic directions. From [29].
For higher temperatures, the Fermi distribution (Eq. 2) extends to higher energies which allows some electrons to occupy the conduction band. This creates a conduction band that is partially occupied by electrons and a valence band that is partially occupied by holes, a hole being the lack of an electron. These two bands can thus contribute to a current and the material conductivity increases the more electron-hole pairs are created. The number density of each of the charge carries, electrons n and holes p, can be determined by an integral over the distribution function (Eq. 2)
multiplied by the density of states for the conduction 𝑔𝑔𝑐𝑐(𝐸𝐸) and the valence band 𝑔𝑔𝑣𝑣(𝐸𝐸) ,
respectively [23]: 𝑛𝑛 = � 𝑔𝑔𝑐𝑐(𝐸𝐸)𝑓𝑓(𝐸𝐸)𝑑𝑑𝐸𝐸 𝐸𝐸𝑡𝑡𝑡𝑡𝑡𝑡 𝐸𝐸𝑐𝑐 Eq. 3 𝑝𝑝 = �𝐸𝐸𝑣𝑣 𝑔𝑔𝑣𝑣 (𝐸𝐸)[1 − 𝑓𝑓(𝐸𝐸)]𝑑𝑑𝐸𝐸 𝐸𝐸𝑏𝑏𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑏𝑏 Eq. 4
For an intrinsic semiconductor, the number density of holes is always equal to the number density of electrons. Taking the density of states derived for an electron in a simplified infinite square well potential, the position of the Fermi level can be calculated [23]:
𝐸𝐸𝐹𝐹=𝐸𝐸𝑐𝑐+ 𝐸𝐸2 𝑣𝑣+3𝑘𝑘4 ln �𝐵𝐵𝑇𝑇 𝑚𝑚𝑝𝑝 ∗ 𝑚𝑚𝑛𝑛∗�,
Eq. 5 where mn* and mp* are the effective masses of holes and electrons respectively. The Fermi
level is thus positioned approximately half-way between the conduction and the valence band, provided the effective masses are similar and the semiconductor is at room temperature (300° C). An additional mechanism for creating charge carriers exists for extrinsic semiconductors. They can be produced by doping which is the addition of small quantities of atoms having a different valency compared to the intrinsic semiconductor. Some materials are inherently extrinsic, among which is the Ta3N5 phase [30].
11 An extrinsic semiconductor has occupied (empty) energy states within its band gap that can donate (accept) electrons to (from) the conduction (valence) band. As these levels usually lie close to the conduction (valence) band, electrons are injected (accepted) already at room temperature, increasing the charge carrier density and rendering the material more conductive compared to the intrinsic semiconductor. As this mechanism either creates only electrons or holes, one of these charge carriers now becomes dominant and is said to be the majority carrier. The material is then
called n-type if electrons dominate, or p-type if holes dominate the electric conduction [23]. Ta3N5
is an n-type semiconductor [30].
At ambient temperature, the extrinsic mechanism dominates over the intrinsic one. The resulting
dominance of one charge carrier to the other directly affects EF, which now becomes [23]:
𝐸𝐸𝐹𝐹= 𝑘𝑘𝐵𝐵𝑇𝑇𝑇𝑇𝑛𝑛(𝑁𝑁𝐴𝐴,𝐷𝐷/𝑛𝑛𝑖𝑖) ± 𝐸𝐸𝐹𝐹,𝑖𝑖 Eq. 6
where EF,i is the Fermi level for the intrinsic semiconductor, ND,A the donor or acceptor density and
𝑛𝑛𝑖𝑖 the intrinsic charge carrier density. The plus applies to the case of an n-type material, the minus
to the case of a p-type material. One can see, that the Fermi energy is shifted up or down from its mid-band position depending on the type of doping. This is illustrated in Fig. 5.
Fig. 5: Schematic illustration of the Fermi level in intrinsic semiconductors, n-type and p-type semiconductors.
2.1.2 Electrolytes
While the conductivity of a metal or semiconductor is provided by electrons, the conductivity of an electrolyte is provided by ions. A water-based electrolyte is made by adding either an acid or base into water. To limit corrosion of electrodes, alkaline solutions are normally preferred in electrolysis [31] and KOH is a typical base used for electrolysis due to its high conductivity in an aqueous
solution [32]. It dissolves in water according to Eq. 7 under the formation of K+ and HO- ions:
𝐾𝐾𝑂𝑂𝐻𝐻𝑎𝑎𝑎𝑎 ⇆ 𝐾𝐾++ 𝐻𝐻𝑂𝑂− Eq. 7
Some of the HO- react with H3O+ present in the solution from the autoionization of water which
increases the pH value of the electrolyte [32]: 𝑝𝑝𝐻𝐻 = − log𝑐𝑐𝐻𝐻3𝑂𝑂+
𝑐𝑐0 = 13.9965 − log 𝑐𝑐0 𝑐𝑐𝑂𝑂𝐻𝐻−
12
In such an alkaline solution, the two redox reactions shown in Eq. 9 and Eq. 10 can principally occur at metallic electrodes held at a certain minimum potential difference of 1.23V. The equations indicate, that hydrogen gas is formed in the forward reaction of Eq. 9 while the backward reaction of Eq. 10 yields oxygen. The former is called the reduction of water, while latter is the oxidation of
HO-. The corresponding electrochemical couples are H2O/H2and O2/HO-, respectively. To prevent
the built-up of charges, the reduction and oxidation have to occur simultaneously and the overall reaction is given by Eq. 1.
2𝐻𝐻2𝑂𝑂 + 2𝑒𝑒−⇌ 𝐻𝐻2+ 2𝐻𝐻𝑂𝑂−, 𝐸𝐸0= −0.83𝑉𝑉 𝑣𝑣𝑣𝑣. 𝑆𝑆𝐻𝐻𝐸𝐸 Eq. 9 𝑂𝑂2+ 2𝐻𝐻2𝑂𝑂 + 4𝑒𝑒−⇌ 4𝐻𝐻𝑂𝑂−, 𝐸𝐸0= +0.40𝑉𝑉 𝑣𝑣𝑣𝑣. 𝑆𝑆𝐻𝐻𝐸𝐸 Eq. 10
𝐸𝐸0 is the standard potential of the corresponding half-cell reaction, which corresponds to the
reversible potential at standard conditions, i.e. an electrode immersed into a solution of 1𝑀𝑀 at 25°𝐶𝐶
[32]. It is measured against the standard hydrogen electrode (SHE), which is a reference electrode to standardize the measurement of electrode potentials of half-cell reactions [32]. Accordingly, the standard potential of the SHE is defined to be at 0.0V [32]. The sum of the two standard potentials from Eq. 9 and Eq. 10 yield the energy required to split water, which is 1.23eV per H atom. The standard potential of a redox couple reflects the electron energy level in an electrolyte. This is
termed Fermi level of the redox couple EF,redox. Its position with respect to EF of a solid electrode is
important as it determines the direction and rate of a charge transfer. For comparing both, the electrochemical scale, measuring a potential against an arbitrarily chosen reference electrode, needs to be converted to the absolute vacuum energy scale that is commonly used in solid state physics. This can be done by knowing that the standard potential of the SHE measured against the potential
of an electron in field-free vacuum 𝐸𝐸𝑎𝑎𝑎𝑎𝑎𝑎0 is estimated to lie at (4,44±0.02)V at 25°C according to a
IUPAC definition [33] [34]. This translates to a Fermi level EF,redox0 of the SHE of
(-4,44±0.02)eV.
Any deviation from the standard state causes EF,redox to shift. Taking a reaction of type
red ↔ ox + e- , where ox stands for the species on the oxidized side of the equation and red for those
on the reduced side of the equation. Then, the variation of EF,redox with the concentration of the
oxidized species cox and that of the reduced species cred is given by the Nernst equation [35]:
𝐸𝐸𝐹𝐹,𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = 𝐸𝐸𝐹𝐹,𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟0 − 𝑘𝑘𝐵𝐵𝑇𝑇 ln �𝑐𝑐𝑐𝑐𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑟𝑟�,
Eq. 11
EF,redox thus changes in particular with the concentration of the HO- ions, which is related to the pH value and Eq. 11 can be expressed as follows [36]:
𝐸𝐸𝐹𝐹,𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 ∝ −59mV/pH Eq. 12
It is important to note, that due to chemical processes at the surface, a similar shift of the band
positions of a semiconductor can be measured, if the HO- are the potential determining ions [37].
This applies in particular to all semiconductor oxides and is confirmed also for Ta3N5 [19]. This
means, that the important difference between the band positions on the surface and EF,redox remains
constant with changes of the pH value. 2.1.3 Interfaces
The photoelectrolysis of water is an interface phenomenon, so that the individual components of a photoelectrochemical cell cannot be regarded separately. Their interaction at the interfaces is important for an understanding of the physics and chemistry of solar water-splitting.
13
Metal-electrolyte interface
An inert metallic electrode put into contact with an electrolyte has, initially, a Fermi level different from the level of the Fermi level of the electrolyte. This is an unstable situation that triggers the transfer of electrons from the higher to the lower level, accumulating a space charge at either side
of the interface. As a consequence, EF of the metal and EF,redox of the electrolyte move towards each
other until they are aligned [36].
Fig. 6: Simplified schematics of the Helmholtz layer at the surface of a metallic electrode that is formed by adsorbed ions on the surface. Drawn after [36].
Fig. 7: The production of hydrogen and oxygen by two metal electrodes held at a minimum potential difference of 1.23V.
A simplified view on the accumulated charges in steady state is depicted in Fig. 6. On the electrode, the space charges are located on the surface as the high conductivity of the metal does not sustain electric fields. The electrolyte forms a more extended space charge region of around 0.1nm that is termed Helmholtz layer [37]. It is made up of solvated ions, specifically absorbed ions and oriented solvent molecules.
In order to produce oxygen and hydrogen from an electrolyte, two electrodes are necessary with a minimum differential potential of 1.23eV (Fig. 7). In this way, electrons can transfer from the
14
negatively biased electrode onto the solution while the other electrode accepts electrons from the electrolyte. In general, the thermodynamically necessary energy of 1.23eV is not sufficient as energy must be provided to compensate resistive losses. In addition, for an appreciable kinetics of the process, a higher potential difference must be envisaged and for the photoelectrolysis, an
overpotential of 0.8V is estimated to be necessary [38]. This translates into a band gap of ≳2.03eV.
Semiconductor-solution interface
Under equilibrium conditions, the physics at the semiconductor/electrolyte interface is similar to that at the metal/electrolyte interface. In contact, the semiconductor Fermi level aligns with the Fermi level of the electrolyte by the transfer of charges. These, again, results in a charge accumulation of opposite polarity on each side of the interface (Fig. 8). Due to the typically much lower charge carrier density in a semiconductor compared to a metal, though, the charges on the
semiconductor side extend deeply into the material. With width 𝑤𝑤𝑎𝑎𝑐𝑐 of this space charge layer
typically ranges from 10nm to 1000nm depending on the charge carrier concentration [35].
Fig. 8: Left: Ta3N5 semiconductor and an electrolyte not in contact. CBM and VBM are the
conduction band minimum and the valence band maximum, respectively. Right: Ta3N5 and the
electrolyte are in contact and in equilibrium.
For an n-type semiconductor such as Ta3N5, the Fermi level EF of the bulk of an unbiased
semiconductor not in contact with the solution normally lies above the electrochemical potential of
the electrolyte EF,redox. The alignment thus forces the built-up of a negative space charge in the
Helmholtz layer and a positive space charge on the semiconductor. The resulting electric field, directed from the semiconductor to the electrolyte, bends the valence and conduction bands upwards, creating a Schottky-like barrier. This separates any electron-hole pairs created in that region. Electrons are pushed towards the bulk and holes are attracted towards the interface. Biasing the semiconductor via an ohmic rear contact, shifts the Fermi level of the semiconductor (Fig. 9). In an ideal case, the band positions at the surface, however, remain fixed [36], so that the band bending is more shallow for a cathodic or forward bias, or steeper for an anodic or reverse bias [35]. The potential at which the bands do not show any bending is termed flat-band potential Vfb.
15
Fig. 9: Left: Cathodic (forward) bias achieved by a applying a negative voltage to the semiconductor. Right: Anodic (reverse) bias from a positive voltage applied to the semiconductor.
Fig. 10: N-type semiconductor in contact with an electrolyte under illumination.
Illuminating the semiconductor disturbs the equilibrium between holes and electrons and their
formerly common Fermi level 𝐸𝐸𝐹𝐹 splits up into separate quasi-Fermi levels for electrons EF,n and
holes EF,p (Fig. 10). The charge separation gives rise to a photovoltage Vphoto that has a similar effect
as the forward bias in Fig. 9. Under illumination EF,n moves upwards above the hydrogen evolution
potential, so that the transfer of electrons from the counter-electrode onto the electrolyte is thermodynamically favorable [35].
On the other hand, EF,p shifts downwards. The significant shift is due to the drastic relative increase
in the number density of holes ∆p compared to the number density under dark conditions
(∆p >> p0). Under illumination therefore, EF,p lies close to the valence band [35] and below the oxygen evolution potential. Holes can then transfer onto the electrolyte to produce oxygen on the semiconductor surface [39]. Under these conditions, water can be electrolyzed according to Fig. 3. 2.1.4 Prerequisites on semiconductors for photoanodes
As can be understood from the discussion above, not all semiconductors can be used as photoanodes. The material requirements are harsh and are listed in the following [39]:
1) Valence and conduction band position
The main prerequisite for a semiconductor electrode to work as a photoanode is that its quasi Fermi levels under illumination straddle the oxygen and hydrogen evolution potentials. In this way the
16
transfer of holes and electrons onto HO- or H2O are energetically favored. This implies that the band
gap is larger than the minimum energy required to split water, i.e. 1.23eV. In practice, as discussed above, overpotentials are necessary that can amount to up to 0.8V so that a minimum band gap of ≳2.03eV is necessary [38].
The straddling condition can be relaxed when using an external bias that allows to lift electrons from the semiconductor to the counter-electrode above the water reduction potential. In this case,
at least the quasi-Fermi level of the holes must be below the oxidation potential of HO-.
While above requirements are critical for the pure functioning of the photoanode, other requirements impact the efficiency of the water splitting [39]:
2) Absorption and band gap
For an efficient charge carrier creation mechanism, the absorption coefficient should be high from the UV range to the band gap. At the same time the band gap should be as small as possible to absorb a large part of the solar spectrum.
3) Charge carrier transport
The efficient transport of charge carriers to the conducting substrate and the semiconductor interface, is necessary to reduce recombination losses. The parameter describing this is the minority
carrier diffusion length 𝐿𝐿𝐷𝐷 or the μτ-product. This parameter is affected by both, the quality of the
semiconductor and the inherent charge carrier properties of the material.
4) Charge carrier injection
Created charge carriers need to transfer across the semiconductor/electrolyte interface sufficiently fast to avoid their accumulation and recombination on the semiconductor. The oxygen evolution reaction on an n-type semiconductor is usually rate-determining as four charges need to be
transferred for the production of one O2 molecule [36]. Enhancing the charge transfer kinetics can
be done by using a co-catalyst such as Co [40], IrO2 [21] [16], Co-phosphite (Co-Pi) [41] or RuO2
[42]. 5) Stability
Finally, the semiconductor needs to be stable in an aqueous solution at elevated pH values. This is typical investigated using XPS [21] to observe changes in the chemical order on the surface of the photoanode. Long-term performance tests, however are rarely found in literature and the studies found are still far away from the 5000h stability tests recommended by leading experts in this field [43].
2.1.5 Semiconductors for photoelectrolysis
Some possible semiconducting materials that fulfill at least the first requirement have been
identified in the past. The most promising and most researched being hematite (Fe2O3)[44], BiVO4
[45] and WO3 [46], Cu2O [47],TaON [48] and Ta3N5 [49]. The pace for identifying new materials
is taken to a new level by the use of computational screening studies, of which two are published to date.
Castelli et al. [50] screened binary oxides of the perovskites type (ABO3) and their derivatives such
as oxinitrides (ABO2N), single and bimetal nitrides (AxNy) and (AxByNz) and single metal
oxinitrides (AxOyNz)[50]. A material was selected when it fulfilled each of three selection criteria:
a band gap from 1.5 to 3.0eV, a negative formation energy and a straddling of the oxygen and the hydrogen evolution potentials at least for the direct band gap. The yield, however remained poor. From the 5400 materials tested, only 15 fulfilled the requirements from which 6 where already known.
17 The study of Wu et al. [51] was somewhat less restrictive, as the crystal structure was not limited
to perovskites. They tested single metal nitrides (AxNy), single metal oxi-nitrides (AxOyNz)and
bimetal oxinitrides (AxByOz1Nz2) based on d0 (Ti4+, Zr4+, Hf4+, V5+, Nb5+, Ta5+, Cr6+, Mo6+, Sc3+, Y3+) and d10 (Ga3+, In3+, Ge4+, Sn4+, Sb5+, Bi5+) cations [52]. These atoms where artificially assembled by taking known structures from the Inorganic crystal structure database (ICSD). Similar to the study by Castelli et al [50], materials were selected based on a band gap with a slighly larger range between 1.3eV and 3.6eV. Their phase stability was determined by measuring the difference in formation energy to a known stable phase. And finally, an empirical law was applied to determine the band positions with respect to the oxygen and hydrogen evolution potential. Similar to the study above, the yield was poor. From the 68 binary nitrides, the 1503 ternary oxinitrides and the 1377 quarterny oxinitrides, in total 27 candidates were identified, 11 of them were already known and
among them was Ta3N5.
It becomes clear, that the requirements for a material to work as a photoanode are harsh and eliminate most semiconductors known to date. This becomes even more pronounced when considering the rather loose criteria applied to select materials with large ranges of allowed parameter values such as the band gap. As all parameters retrieved for those unknown phases are based on simple (semi-)empirical models, the ranges need to allow for uncertainties, though. It is thus clear that the already small yield from those studies may become even smaller when measurements on experimentally prepared materials are known.
2.1.6 Efficiencies and benchmarks
Different definitions to benchmark the solar water-splitting efficiency are used to compare device performances. The most general form is the solar-to-hydrogen efficiency which is simply the ratio
between the energy of the hydrogen gas that is produced per time (jphotoVredox) and the incoming
solar power Plight. The first still needs to be corrected by the supplementary power input from a
potential bias voltage with respect to the reversible hydrogen electrode RHE (jphotoVbias, RHE).
η𝑆𝑆𝑆𝑆𝐻𝐻 =𝑗𝑗𝑝𝑝ℎ𝑟𝑟𝑜𝑜𝑟𝑟�𝑉𝑉𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟− 𝑉𝑉𝑎𝑎𝑖𝑖𝑎𝑎𝑎𝑎,𝑅𝑅𝐻𝐻𝐸𝐸� 𝑃𝑃𝑙𝑙𝑖𝑖𝑙𝑙ℎ𝑜𝑜
Eq. 13
where jphoto is the photocurrent, Vredox is the minimum redox potential to split water, Vbias is the bias
voltage applied between the reference and the semiconducting electrode. Vredox can be taken
as 1.23eV, corresponding to the Gibbs free energy change of 238kJ/mol. From Eq. 13, theoretical limits for solar-water-splitting as a function of the band gap of a material can be calculated
(Fig. 11). Taking the Ta3N5 semiconductor with a band gap of 2.1eV, i.e. providing for an
overpotential of roughly 0.8V, results in a theoretical maximum solar-to-hydrogen efficiency of 15.9% [53].
18
Fig. 11: Theoretical limits for solar water splitting as a function of the band gap. The optimum band gap is determined to be around 2eV. From [38].
To take the band gap dependency on the efficiency out of the equation, the incident-photon-to-current efficiency is used. It gives the efficiency of a photon to create a charge carrier that contributes to the photocurrent, i.e. recombined charge carriers are not take into account:
𝐼𝐼𝑃𝑃𝐶𝐶𝐸𝐸(𝜆𝜆) =𝜆𝜆𝜆𝜆ℎ𝑐𝑐𝑗𝑗𝑃𝑃𝑝𝑝ℎ𝑟𝑟𝑜𝑜𝑟𝑟(𝜆𝜆) 𝑙𝑙𝑖𝑖𝑙𝑙ℎ𝑜𝑜(𝜆𝜆),
Eq. 14
where h is Planck’s constant, c is the speed of light, λ is the wavelength and q the elementary charge.
Some benchmarks are published and listed in Table 4 for Ta3N5 photoanodes. In general, the overall
efficiency still remains small with a solar-to-hydrogen efficiency barely exceeding 1%. Using different configurations than the single semiconductor photoanode described above, higher efficiencies can be reached. However, this comes at the cost of added complexity. One example is
the use of a heterojunction tandem absorber combined with a Rh-functionalized, AlxIn1-xP
passivated surface and an RuO2 counter-electrode that holds the current record in STH efficiency
for a bias-free photoelectrolysis of 14% [15].
Table 1: Published benchmarks for Ta3N5-based photoanodes.
Photoanode η𝐒𝐒𝐒𝐒𝐒𝐒, % Ref.
Ta3N5 nanorods with IrO2 co-catalyst 0.5% [16]
[17] Ba-doped Ta3N5 nanorods with Co-Pi
co-catalyst
1.56% @ 0.87V vs. RHE [17]
Ba-doped Ta3N5 nanorods without
cocatalyst
0.89% @ 0.93V vs. RHE [17]
19
The Ta
3N
5crystalline phase
2.2.1 Crystalline phases of the Ta-O-N ternary system
Tantalum belongs to the transition metals, which are known for their variety of oxidation states [54] and the stability of their compounds over large non-stoichiometric ratios [55]. Of interest for this
thesis are in particular its compounds formed with nitrogen (Chapter 2.2.1.1). But as Ta3N5 is only
synthesized so far by incorporating oxygen into the lattice, some important oxygen-containing phases are discussed as well (Chapter 2.2.1.10). Finally, formation enthalpies of some phases are discussed as they offer valuable clues on the stability of certain phases (Chapter 2.2.1.3).
2.2.1.1 The Ta-N binary system
The Ta-N system knows 9 crystalline phases from the pure tantalum phase and the solid solution
Ta(N) up to the strongly nitrided Ta3N5 phase (Table 2). Most of them are metallic phases without
interest for this thesis. It is however, instructive to follow the changes in chemistry, structure and the optical and electronic properties when successively adding nitrogen to an initially pure Ta crystal. Also, it is important to define the phases, as names can change between publications. The δ-TaN phase in [56], for example, is a hexagonal phase with lattice parameters a=0.293nm and c=0.289nm and of WC-type, while in another publication, the δ-TaN phase has a rocksalt structure with lattice parameter a=0.43nm [57]. The definition of phases used in this thesis is based on the
PhD thesis by Ganin from 2005 [58] and is complemented by the Ta2N3 phase [59].
Pure Ta crystallizes in the body-centered cubic (bcc) structure and can accomodate interstitial N up to a 4at.% to form the solid solution Ta(N) [60]. At higher nitrogen at.%, the Ta atoms arrange in a closed-packed hexagonal structure to accommodate N in every second octahedral hole [56]. This is a typical phase transition that can be observed in many transition metal nitrides and also carbides [56]. In the Ta-N system, the homogeneity region of this phase is centered around 30at.% (Fig. 12) and is termed β-Ta2N.
Fig. 12: Ta-N phase diagram (from [58] [61]).
The TaN phase exists in three polymorphs, the rocksalt δ-TaN, the hexagonal ε-TaN (CoSn-type) and the hexagonal θ-TaN (WC-type). Latter is the most stable structure, followed by δ-TaN and ε-TaN being the least stable structure judging from DFT calculations [62].
δ-TaN is a high-pressure, high-temperature phase that can accommodate excess nitrogen far beyond
its stoichiometry. TaNx phases were prepared with x going up to 1.66 [63] which exceeds even the
N at.% of Ta3N5. The nonstoichiometry in δ-TaN is accompanied by Ta vacancies, that, with
20
Table 2: Summary of known stoichiometric phases of the binary system Ta-N. Phases and reference were taken from [58] and complemented. For some phases, alternative structures exist, that are not mentioned in this table. The bct-TaN phase is not listed in the Inorganic crystal structure database (ICDS), but was found by magnetron sputtering by two different groups [55] [64] and was later identified as Ta2N3 [58].
phase chemical
formula
crystal structure and lattice constants (nm)
space group ref.
Ta Ta bcc (W-type)
a = 0.3311 𝐼𝐼𝑚𝑚3�𝑚𝑚
[56]
α-Ta(N) TaN0.0 -TaN0.04 bcc, solid solution
a = 0.3315 𝐼𝐼𝑚𝑚3�𝑚𝑚
[60] β -Ta(N) TaN0.05 bcc, α phase with 3x3x3
superstructure a = 10.11 (3x3.369) 𝐼𝐼𝑚𝑚3�𝑚𝑚 [56] β-Ta2N (γ-TaN) TaN0.41 TaN0.5 hexagonal a = 0.3041; c = 0.4907 a = 0.3048; c = 0.4919 𝑃𝑃3�𝑚𝑚1 [56] [60] δ-TaN TaN0.91 TaN0.99 TaNx fcc (NaCl type) a = 0.4337 a = 0.4336 a = 0.498 to 5.047 𝐹𝐹𝑚𝑚3�𝑚𝑚 [57] [65] [55] ε-TaN TaN1.00 TaN1.00 TaN1.00 hex. (CoSn-type) a = 0.5185; c = 0.2908 a = 0.5191, c = 0.2911 a = 0.5188; c = 0.2903 𝑃𝑃6/𝑚𝑚𝑚𝑚𝑚𝑚 [56] [60] [66] θ-TaN TaN1.00 TaN1.08 hex. (WC-type) a = 0.2936; c = 0.2885 a = 0.2938; c = 0.2868 𝑃𝑃6𝑚𝑚2 [67] [57] Ta2N3, (bct-TaN)
Ta2N3 cubic (La2O3-type)
a = 0.98 𝐼𝐼𝐼𝐼3� [59] [55] [64] Ta5N6 Ta5N6 hexagonal a = 0.5175; c = 1.0307 𝑃𝑃63𝑚𝑚𝑐𝑐𝑚𝑚 [55]
Ta4N5 Ta4N5 tetragonal (defective NaCl type)
a = 0.6835; c = 0.4272
𝐼𝐼4/𝑚𝑚
[66]
Ta3N5 Ta3N5 orthorhombic (Fe2TiO5-type)
a = 0.38862 b = 1.02118 c = 1.02624
𝐶𝐶𝑚𝑚𝐶𝐶𝑚𝑚
[68]
This can be observed experimentally by an increase of resistivity with increasing the N at.% [69].
A clear border for this metal-to-insulator transition is given in [29], which states that films of Ta1.12N
up to Ta0.74N are metallic and barely metallic respectively while Ta0.66N is an insulator. Electron
counting for a Ta0.6N structure confirms this: In total, 40 valence electrons from the atoms (5 valence
electron from 3 Ta atoms plus 5 valence electrons from 5 N atoms) need to be distributed over the
21 atoms and can thus accommodate 40electrons (2 electrons per s-orbital plus 6 per p-orbital for each of the 5 N atoms). Hence, all valence electrons of the atoms can be accommodated in the valence band of the structure. The conduction band therefore remains empty and the material is an insulator [69].
Four higher ordered nitrides of the Ta-N exists, that are, in order of increasing nitrogen at.%, the
hexagonal Ta5N6, tetragonal Ta4N5 and the orthorhombic Ta3N5 phase. Despite their elevated
content of nitrogen, only the highest nitride, Ta3N5 is a semiconductor, all other phases being
metallic [62]. As Ta3N5 and Ta0.6N (see above) contain the same at.% of N, the same electron
counting example from above applies here. The electronic, optical and electrochemical properties of Ta3N5 are discussed in more detail below.
2.2.1.2 Important phases of the ternary Ta-O-N system
Ta3N5 has only been synthesized so far by the incorporation of oxygen into the lattice, the reason
for which will be discussed later (Section 5.1). Therefore, some important phases containing oxygen are introduced here in addition to those of the Ta-N system summarized in Table 2.
Of importance here is the highly stable orthorhombic Ta2O5 phase (Table 3). It has a large
band-gap of 3.9eV which is responsible for its transparent appearance [19].
Another important phase is the TaON phase. Initially, this chemical formula was believed to have two polymorphs termed α-TaON and β-TaON (Table 3). Recent investigations suggest, however, that the α-TaON phase does not exist and is recommended to be removed from crystallographic databases [70].
TaON belongs to the large group of transition metal oxynitrides. Of particular interest for this class of material is their band gap that can be tuned by changing the O/N ratio. In these materials the conduction band minimum is formed by the empty metal d orbitals while the valence band maximum is formed by the 2p oxygen or nitrogen orbitals. As the nitrogen orbitals are shallower than the oxygen ones, a substitution of oxygen by nitrogen can decrease the band gap, thus changing the color of the material [52]. The same reasoning applies for explaining the much smaller band gap
of Ta3N5 compared to that of Ta2O5. Fig. 13 schematically illustrates the formation of valence and
conduction bands around the band gap.
Table 3: Summary of important phases from the Ta-O-N ternary system.
phase chemical
formula
crystal structure, lattice constants (nm)
space group ref.
α-TaON TaON likely an erroneous entry in literature, see [71]
β-TaON TaON Monoclinic (baddeleyite)
a = 0.494941 b = 0.501662 c = 0.516430 P21/C4 [72] Ta2O5 Ta2O5 Orthorombic a=0.6198 b=4.290 c=0.3888 Pccm [73]
22
Fig. 13: The formation of band structures by bonding of a transition metal to oxygen and nitrogen. The top of the valence band are formed by O 2p orbitals while the bottom of the conduction band is formed by empty transition metal 5d orbitals. The schematics illustrate this at the example of the pure oxide NaTaO3 (left) and the oxynitride BaTaO2N. From [52].
2.2.1.3 Formation enthalpies of some selected phases from the Ta-O-N ternary system
The knowledge of the stability of a phase is important for its synthesis, as stable materials are usually easier to synthesize than metastable materials. The energetics of the binary Ta-N system is shown by help of a convex hull plot in Fig. 14. It is a scatter plot of formation enthalpy vs. anion molar fraction and by connecting the structures having the lowest (most negative) formation enthalpies the stable phases can be distinguished from the metastable ones [74]. It can be seen from this plot, that the structures Ta2N and Ta5N6 and Ta3N5 are stable, while the phases Ta4N5, ε-TaN and δ-TaN are metastable. This result was confirmed independently by a different group [75]. An interesting feature on the convex hull plot in Fig. 14 is the evolution of formation enthalpy for
the non-stoichiometric δ-TaN phase (TaN cubic, TaN1.13, TaN1.33). As the stoichiometry approaches
the Ta3N5 stoichiometry, the formation enthalpy of the δ-TaNx phase diminishes. The data point at
which the two phases have an identical stoichiometry (δ-TaN1.67 vs. Ta3N5) is not shown in this
study. The question which of the two phases has the lower formation enthalpy at this point, can therefore not be answered.
Values of experimentally measured formation enthalpy for some phases of the ternary Ta-O-N
system are summarized in Table 4. It can be seen that, Ta3N5 has, as in Fig. 14, a very low formation
enthalpy of -850 kJ/mol. Calculations show in addition, that the substitution of N by O can lower the formation enthalpy even more [76] [77] [78].
Table 4: Experimentally measured formation enthalpies of selected phases from the Ta-O-N system measured under standard conditions 298.15 K, 0.1MPa [79].
phase formation enthalpie 𝛥𝛥H (kJ/mol) ref.
Ta 0.0 (by definition)
Ta3N5 -849.7 ± 11.7 [80]
TaON -635.51 ± 4.27 [76]
23
Fig. 14: Formation enthalpies as calculated by DFT for some phases of the Ta-N system. The hexagonal TaN phase is the ε-TaN phase in Table 2 and the nonstoichiometric phases TaN1.13 and
TaN1.33 are all derived from the δ-TaN phase (from [82]).
2.2.2 The crystalline Ta3N5 phase
This subchapter serves to summarize known properties of the Ta3N5 phase that are relevant for the
water splitting process: structure, optical properties (band gap, dielectric function), electrical properties (carrier mobility) and electrochemical properties (band position, stability).
2.2.2.1 Ta3N5 structure
Ta3N5 crystallizes in an orthorhombic structure with the space group
𝐶𝐶𝑚𝑚𝐶𝐶𝑚𝑚 (Fig. 15). In this crystal, Ta is 6-fold coordinated by N and the TaN6 octahedra are
irregularly oriented. Each N is three or four-fold Ta coordinated and this structure allows an
increased accommodation of N compared to rocksalt δ-TaN that has oriented TaN6 octahedra, with
each N being 6-fold Ta coordinated (Fig. 15) [18].
Fig. 15: Ta3N5 crystal: unit cell. The grey spheres indicate the position of the Ta atoms, the blue
ones the positions of the N atoms. From [18].
There is still some confusion in the literature, whether the structure is orthorhombic as stated in [83] or monoclinic as proposed in [66] (cf. e.g. [18] [20] for references to the orthorhombic structure and e.g. [84] [40] for reference to the monoclinic structure). In both references, the lattice constants are determined to be a = c = 1.0264nm compared to b ≈ c ≈ 1.03nm and b = 0.3893nm compared to a = 0.37nm for the orthorhombic structure and the monoclinic one, respectively. The determining parameter is thus the angle which is stated to be β=90° in the case of the orthorhombic interpretation
24
and β≈90° in the case of the monoclinic interpretation. The proposed structures are thus at least very
close to each other. The latest experimental results on the structure of Ta3N5 is obtained by Brese
et al. [68], using neutron diffraction. They come to the conclusion that both structures are too close to each other to discern by their experiemnt, but decide that the orthorhombic structure is likely the correct one. In this thesis, the structure refined by Brese et al., referenced under the powder diffraction file (PDF) 01-079-1533, is thus used.
2.2.2.2 Ta3N5 band gap and dielectric function
Ta3N5 is an indirect semiconductor as can be seen from the band structure in Fig. 16. Its direct band
gap is often stated to lie around 2.1eV with an indirect band gap below that value [18], possible around 1.9eV [85]. The optical properties that result from such a band structure can be calculated from the complex dielectric function. Two measurements exits to date using ellipsometry.
The first one was published by Langereis et al. [86]. They deposited a thin film of around 50nm of
Ta3N5 by atomic layer deposition (ALD) and measured the dielectric constant by ellipsometry
in-situ. The second study, published by Morbec et al. [18], was a computational study that was validated using experiments. The sample was prepared by depositing 29nm of Ta by electron beam
evaporation. The film was then oxidized at 700°C under a flow of O2 for 1h and subsequently nitride
at 850°C in NH3 for 8h. The results from both studies are plotted in Fig. 16.
A comparison reveals that both results do not match very well. Some difference can be explained by the different preparation methods as well as that the measurement of Langereis et al. was done in-situ [86]. As the dielectric function in the measurement of Morbec et al. reveals more features, it likely is more accurate while the features in Langereis et al. can be imagined to be smeared out.
Fig. 16: Left: Band diagram of Ta3N5 calculated by DFT. The blue arrow indicates the indirect
band gap. It is smaller than the usually stated 2.1eV which is an intrinsic shortcoming of conventional DFT calculations. From [18]. Right: Complex dielectric function measured by Langereis et al. [86] and Morbec et al. [18]. The two vertical line at 2.1eV indicates the position of the direct band gap.
Using some simple conversions between the dielectric function and the refractive index N = n + jk (cf. Section 3.4), the absorption coefficient α = 4πk(λ)/λ can be calculated. Its inverse is the
25 absorption length, which is the thickness of an absorbing medium that decreases the initial intensity
I0 of an incident beam to 1/3 ≈ 37%. Both, the absorption coefficient and its absorption length are
plotted in Fig. 17. The figures show that despite the band gap of 2.1eV, a considerable absorption for thin films of around 100nm only occurs for photons having energies greater than around 2.7eV. Below that, the film needs to considerably thicker by a factor of 10 to increase absorption between 2.7eV and 2.2eV.
Fig. 17 : Absorption coefficient α and absorption length for the measured dielectric function by Morbec et al. [18] and Langereis et al. [86].
2.2.2.3 Ta3N5 charge carrier mobility
The absorption is naturally in competition with the electronic properties as the first demands a film as thick as possible and the latter demands it to be as thin as possible to avoid recombination of excited charge carriers. The parameter to describe the charge carrier dynamics is the
mobility-lifetime µτ product that describes, how fast a particle propagates and on what timescales it
recombines. Using the µτ-product of the minority carriers, the minority carrier diffusion length LD
can be calculated [87].
𝐿𝐿𝐷𝐷= �𝑘𝑘𝐵𝐵𝜆𝜆𝑇𝑇(𝜇𝜇𝜇𝜇)𝑚𝑚𝑖𝑖𝑛𝑛𝑟𝑟𝑟𝑟𝑖𝑖𝑜𝑜𝑚𝑚, Eq. 15
where kB is the Boltzmann constant, T the temperature and q the elementary charge.
Theoretical studies by Morbec et al. [18] and Nurlaela et al. [88] both predict strong anisotropies in
Ta3N5 with some directions having large effective masses for holes and electrons (Table 5). They
suggest this poor bulk charge transport properties to be the reason for the low photocurrents
observed on Ta3N5. More recent experimental measurements, show however, while the mobility is
low, the carrier lifetime is comparable to that of Si (~ ms) which results in a long diffusion length [47].
26
Table 5: Hole and electron effective masses for Ta3N5 in comparison with those of monocrystalline Si.
phase type of study holes 𝒎𝒎𝒉𝒉∗/𝒎𝒎𝟎𝟎 electrons 𝒎𝒎𝒆𝒆∗/𝒎𝒎𝟎𝟎 ref.
Ta3N5 DFT 0.27-3.56 0.23-2.70 [18]
Ta3N5 DFT 0.85-3.38 0.60-1.94 [88]
for comparison:
monocrystalline Si n/a 0.81 1.182 [23]
2.2.2.4 Ta3N5 band positions
The band positions of the Ta3N5 semiconductor have been measured by an electrochemical (EC)
method and by ultraviolet photoelectron spectroscopy (UPS) for a sample prepared by ammonolysis by Chun et al. [19].
Both measurements, conducted by the same authors, result in the same band position. The valence band was found to be at -6.0eV and the conduction band at -4.0eV measured on the absolute vacuum scale at a pH value of 0 [19]. As the oxygen and hydrogen evolution potentials are at -4,5eV and -5,73eV at pH = 0, the measurements confirm that the bands actually straddle the oxygen and hydrogen evolution potential. In the same paper, it is furthermore confirmed, that the band positions
change with pH at the same rate as that of the electrolyte (60mV/pH), so that the relative positioning
of the bands with respect to the oxygen and hydrogen evolution potential remains constant at all pH values.
Fig. 18: Band positions of Ta2O5, TaON and Ta3N5 as measured by ultraviolet photoelectron
spectroscopy (UPS) and electrochemically (EC). From [19].
As these are experimental measurements, the prepared Ta3N5 photoanode likely contains defects
and impurities. It is actually those deviations from the ideal structure that places the bands into the right position as shown by DFT calculations by Harb et al. [78] [89]. They show that the bands of
an ideal Ta3N5 structure lie much higher than depicted in Fig. 18 so that the valence band lies above
the oxygen evolution potential. Only by oxygen incorporation into the lattice (Ta2.83N4.16O0.83), the
bands shift sufficiently downwards so that they bracket the hydrogen and oxygen evolution potential [78]. A different paper by the same author shows that the same downward shift of bands can also be induced by Ta and/or N vacancies [89].
